Modelling financial return distributions: a neat solution for moderately leptokurtic data